> ## Documentation Index
> Fetch the complete documentation index at: https://elephas.app/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# See What Was Hidden in an Answer (Privacy Shield)

> Use the shield under an Elephas answer to see which personal details were replaced with placeholders before your question and sources went to cloud AI.

When personal information protection runs on an answer, a shield appears under it with the number of personal details Elephas hid before sending. Click the shield to see the list, on your computer.

## Before you start

* You are using **Elephas AI** or **your own API key**. Local AI answers have no shield, because nothing is sent.
* **Protect personal information when using cloud AI** is on in **Settings › AI**. See [Personal information protection](/docs/privacy-and-security/personal-information-protection).

## Steps

<Steps>
  <Step title="Ask a question">
    Ask a question in a chat as usual. When the answer is ready, look at the row of actions under it.
  </Step>

  <Step title="Find the shield">
    Next to the sources count (for example **3 sources**) you see a shield with a count, such as **3 details hidden**, **1 detail hidden**, or **Nothing to hide**.
  </Step>

  <Step title="Open the list">
    Click the shield. The **Hidden before sending** panel lists each hidden detail once, with its kind (Name, Organisation, Email, and so on) and the value from your documents. The first five are shown. Click **Show all** to see the rest.
  </Step>

  <Step title="Close the panel">
    Click **Done**, or press Escape or Return.
  </Step>
</Steps>

The shield works the same way on Mac and Windows. It is a button you can reach with the keyboard, and the count is always written on it, so you never need to hover.

## How it works

The panel says: "Elephas replaced these with placeholders before anything was sent to the AI, and put the real words back in the answer on this computer."

**What the count means.** It is the number of different details, not the number of replacements. If a name appears ten times across several requests, it counts once.

**What the list covers.** Everything hidden while Elephas produced that answer: your question, the chat history that was sent, and every passage the AI was given, including passages the final answer did not use. So it is not "the personal details in this answer". It is what was kept out of the requests.

**"Nothing to hide".** Elephas checked your question and its sources before sending and found nothing that needed hiding. This is different from no shield at all, which means hiding did not run.

**Some values are masked in the panel.** Card numbers, bank accounts (IBAN), social security, Aadhaar and PAN numbers, financial details and anything listed as **Other detail** show only their last four characters. Passwords show no characters at all. Everything else is shown as it appears in your files, because you are looking at your own documents on your own computer. At most 200 details are listed; if more were hidden, the panel says how many more.

**It may include things that are not personal.** The detector sometimes hides words that turn out not to be personal information. They are listed because they were hidden.

**When there is no shield:**

| Situation | Shield |
| - | - |
| Elephas AI or your own API key, protection on | Yes |
| Team workspace whose AI policy requires redaction | Yes |
| Protection turned off | No |
| Team workspace whose AI policy allows cloud AI without redaction | No |
| Local AI | No (nothing is sent) |
| Elephas asks you a question back, such as which document you mean | No |
| An answer about Elephas itself given without any AI | No |
| A request that failed or was stopped | No |
| Answers from an earlier session, or after you switch workspace | No |

### The list is kept in memory only

The panel reads: "Kept only until you switch workspace or quit Elephas."

* The list is never saved to the database. After you restart Elephas, earlier answers show no shield.
* It is never synced. Your chat syncs with [Personal Sync](/docs/sync-and-devices/personal-sync-overview), but the list is not part of it, so the same answer on another computer shows no shield.
* It is never written to the log. The log records only how many details were hidden.

This is deliberate. Keeping a second copy of exactly the words the feature protects would create a new place for them to leak.

## Troubleshooting

<AccordionGroup>
  <Accordion title="I don't see a shield under my answer">
    Check which AI answered (the badge reads **Answered by:**). Local AI answers have no shield. Then check that **Protect personal information when using cloud AI** is on in **Settings › AI**. Answers from before your last restart, or from before you switched workspace, also show no shield.
  </Accordion>

  <Accordion title="A detail I expected to be hidden is not in the list">
    Detection hides many common personal details, but not all of them. To make sure a specific name or term is always hidden, add it to the Brain's [Protected Names](/docs/brains-and-documents/protected-names). For the most sensitive material, use [Local AI](/docs/ai-options/run-local-ai), which sends nothing.
  </Accordion>

  <Accordion title="Something harmless was hidden">
    The detector is cautious and sometimes hides words that are not personal. The real words are put back in the answer on your computer, so the answer still reads normally.
  </Accordion>
</AccordionGroup>

## Related articles

<CardGroup cols={2}>
  <Card title="Personal information protection" href="/docs/privacy-and-security/personal-information-protection" />

  <Card title="What data is sent to AI" href="/docs/privacy-and-security/what-data-is-sent-to-ai" />

  <Card title="Protected Names" href="/docs/brains-and-documents/protected-names" />

  <Card title="Run Local AI" href="/docs/ai-options/run-local-ai" />
</CardGroup>


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.